I've now been using the Bayesian mailet for several days now. It's
catching about 99% of the spam and giving me about 1% false-positives.
I realize with continued training over time, the errors should approach
0%. But as a mathmetician, I know they'll never hit absolute 0%. This
leaves me with a problem that I can either just not look at the flagged
spam and cut my losses if I miss an important email that was incorrectly
flagged. Or I can continue to download and scan the hundreds of spams
daily "just to be sure". I run a business, and I can think of a lot of
bad things that will happen if I miss even one important email from a
client.
A whitelist that would bypass the spam checker altogether would help
minimize the risk. But I don't want to manually maintain a whitelist.
Which brings me to the question... It seems reasonable to have a mailet
that simply stores in a db table the target email address(es) of every
outbound note sent by a validated SMTP user, and then compare inbound
notes against this dynamic whitelist. At least if I've ever sent a note
to somebody, I have a 100% guarantee that replies or subsequent notes
will never be inadvertently flagged as spam.
I figure this isn't going to be rocket science to write both the
outbound mailet that stores in the db and the inbound matcher that
matches against the entries in the table. But I would like some
comments on a) if this has already been done with existing
matchers/mailets already available, and b) if there are horribly bad
issues with doing something like this that I haven't thought about? I
realize it will become a very long table in the db. But indexed
searches should still be reasonably efficient.
Again, just curious if this has been beat around before and if so, what
were the results?
Thanks.
Jerry
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
